IIT Delhi

Mechanical Engineering Department

Developing a Viscoelastic creep model for

an epoxy based on experimental data

Supervisor:
Professor Naresh V. Datla (PhD)

Tesfaye O. Terefe and Akash Deep

 Introduction
• Application of polymers in engineering applications
• Cylinders manufactured from polymers to transport gas or oil
• Failures due to high pressure or temperature resulted in
leakage after some years
• Due to extra temperature the pressure at the joint of the
cylinders will reduce, creep strain developed
• Developing the viscoelastic model (time dependent behavior
of material’s response) is very important to design high
pressure gas or oil.

19/05/2021 Tesfaye O. Terefe and Akash Deep 2

 Objective
Aim
• Predicting the viscoelastic behavior of epoxy under creep
test using experimental data
Objectives
• Determination of Prony coefficients for the creep test
experimental data using curve fitting
• Predicting the viscoelastic behavior of an adhesive based
on epoxy resin Epidian53 under creep test using FE
analysis
• Comparison of the experimental data and the FE analysis
result

19/05/2021 Tesfaye O. Terefe and Akash Deep 3

 Methodology
Literature Review

Determination of prony series coefficients

• Using Matlab 2016a from the stress strain relationship

• Maxwell prony series model is used

Dogbone geometric modelling

• Abaqus 6.14 (SV) is used to develop the geometry

FE analysis

• Abaqus 6.14 (SV) is used for FE simulation

• Based on the Maxwell model the coefficients determined from the experimental
data by curve fitting is used to model the viscoelastic material in Abaqus 6.14

Comparison of experimental and FE analysis result

• Comparison between the data extracted from the experimental and FE analysis is
done

19/05/2021 Tesfaye O. Terefe and Akash Deep 4

 Materials and methods
• Experimental data
• Epoxy: Epidian53:PAC
• Dogbone specimen model
• FE analysis (Abaqus 6.14 Student Version)
• Establishment of the viscoelastic property is done in
Abaqus 6.14 student version

Fig: Dog bone Specimen

19/05/2021 Tesfaye O. Terefe and Akash Deep 5
 Viscoelastic models: Maxwell Model
• It is the most widely used method to capture the behavior of
relaxation of polymers within a concise time variation.
• It consists of a linear spring element and linear viscous
dashpot element connected in series.

19/05/2021 Tesfaye O. Terefe and Akash Deep 6

 Experimental data
• Predetermined data used from literature

19/05/2021 Tesfaye O. Terefe and Akash Deep 7

 Cont.
• Experimental data strain
0
stress
0
strain
0
stress
0
strain
0
stress
0
0.032195 0.526529 0.006559 -0.21923 0.078138 0.175903
0.122792 1.798955 0.162176 1.492154 0.20156 1.185352
0.277779 4.475244 0.564352 5.661025 0.39639 2.853087
0.477965 7.941253 1.005424 10.26871 0.87719 6.627653
0.67815 11.40726 1.342781 13.69161 1.241012 9.524384
0.891339 14.96106 1.45297 14.96414 1.494359 11.58718
1.216838 16.5418 1.870011 16.36994 1.74129 13.4745
1.542336 18.12253 2.254485 17.64398 1.91014 14.92279
1.8483 19.61543 2.717212 19.09388 2.32088 15.97767
2.134901 20.75735 2.990895 20.01643 2.725203 16.85708
2.499726 22.119 3.544878 21.73 3.083841 17.69238
2.916653 23.70024 4.001131 23.09215 3.494638 18.65955
3.307571 25.10589 4.437793 24.45419 3.827212 19.40698
3.64003 26.02876 4.665905 25.1572 4.081503 20.02241
4.116075 27.08401 5.018213 25.68544 4.518566 20.77042
4.474714 27.91931 5.500933 26.52142 4.903498 21.34271
4.85936 28.93019 5.931579 27.09396 4.903498 21.34271
5.387535 30.16115 6.466457 28.0618 5.320998 22.04675
5.855933 22.92688
@ t=0.15hr 6.903577 28.7221
6.403873 23.89479
7.360203 29.51407
7.810412 30.13058 6.840992 24.55509
7.232398 25.21513
@ t=0.55hr
19/05/2021 Tesfaye O. Terefe and Akash Deep @ t=2.25hr 8
 Curve fitting: Matlab 2016a
• Maxwell model with n=2 is used to determine the prony series coefficients
G(t) = (G1*(1-exp(-G1*t/T1))+G2*(1-exp(-G2*t/T2)))
• Coefficients are determined as:
Coefficients (with 95% confidence bounds):
G1 = -0.02936 (-0.04575, -0.01298)
G2 = 0.02971 (0.01285, 0.04657)
T1 = 1.644 (0.6598, 2.628)
T2 = 1.393e+05 (-1.857e+09, 1.857e+09)
• Goodness of fit:
SSE: 15.71
R-square: 0.9915
Adjusted R-square: 0.9898
RMSE: 1.023

19/05/2021 Tesfaye O. Terefe and Akash Deep 9

 Cont.
• General model:
G(t) = (G1*(1-exp(-G1*t/T1))+G2*(1-exp(-G2*t/T2)))
• Coefficients (with 95% confidence bounds):
G1 = 0.03063 (0.01581, 0.04544)
G2 = -0.03043 (-0.04491, -0.01595)
T1 = 1.791e+05 (-1.701e+09, 1.701e+09)
T2 = 2.641 (1.372, 3.909)
• Goodness of fit:
SSE: 12.11
R-square: 0.994
Adjusted R-square: 0.993
RMSE: 0.8201
19/05/2021 Tesfaye O. Terefe and Akash Deep 10
 Cont.
• General model:
G(t) = (G1*(1-exp(-G1*t/T1))+G2*(1-exp(-G2*t/T2)))
• Coefficients (with 95% confidence bounds):
G1 = -0.02606 (-0.03813, -0.01398)
G2 = 0.02558 (0.01319, 0.03796)
T1 = 2.496 (1.33, 3.662)
T2 = 1.648e+05 (-1.538e+09, 1.539e+09)
• Goodness of fit:
SSE: 8.268
R-square: 0.9944
Adjusted R-square: 0.9935
RMSE: 0.6597

19/05/2021 Tesfaye O. Terefe and Akash Deep 11

 FE analysis in ABAQUS 6.14 (student
version)
• Geometric modelling

Fig: Dog bone Specimen

• Boundary Conditions
One end: Fixed
Other end: Force (Pressure) is applied for time t
• Meshing element used: 3D Stress element

19/05/2021 Tesfaye O. Terefe and Akash Deep 12

 FE Results
• Based on the viscoelastic modelled in the Abaqus the strain vs
time result obtained for different loading

Fig. Simulation in Abaqus

19/05/2021 Tesfaye O. Terefe and Akash Deep 13
 Strain vs time
• Comparison between the experimental data and FE analysis

Fig. Experimental result Fig: FE simulation result

19/05/2021 Tesfaye O. Terefe and Akash Deep 14
 Conclusion

• From both analysis it is shown that the level of stress has

significant on the model parameters
• Further the FE analysis can be extended to multiaxial creep
behaviour

19/05/2021 Tesfaye O. Terefe and Akash Deep 15

 References
[1]. Pacheco JEL, Bavastri CA, Pereira JT. Viscoelastic Relaxation Modulus Characterization Using Prony
Series. Latin American Journal of Solids and Structures 2014;12.
[2]. Xu Q, Engquist B. A mathematical and Physical Model Improves Accuracy in Simulating Material
Relaxation Modulus and Viscoelastic Responses.
[3]. Alejandro T-RM, Edmundo L-UL. Prony series calculation for viscoelastic behavior modeling of structural
adhesives from DMA data. IngenIería InvestIgacIón y tecnología. 2020.
[4]. Adel WM, Guo-zhu L. Developing a Viscoelastic Relaxation Model for AP-HTPB Composite Solid
Propellant Based on Experimental Data. 21st AIAA International Space Planes and Hypersonics
Technologies Conference; China: International Space Planes and Hypersonic Systems and Technologies
Conferences; 2017.
[5]. Fan P, Chen W, Hu J, Zhao B, Gao J, Fang G, et al. Stress relaxation properties of an epoxy-based shape-
memory polymer considering temperature influence: experimental investigation and constitutive
modeling. Mech Time-Depend Mater 2020.
[6]. Hu WG, Guan HY. Study on compressive stress relaxation behavior of beech based on the finite element
method. 2019.
[7]. Majda P, Skrodzewicz J. A modified creep model of epoxy adhesive at ambient temperature. International
Journal of Adhesion & Adhesives. 2009;29.
[8]. Papanicolaou GC, Zaoutsos SP. Viscoelastic constitutive modeling of creep and stress relaxation in
polymers and polymer matrix composites. Creep and Fatigue in Polymer Matrix Composites. 2019.
[9]. Xu Q, Engquist B, editors. A mathematicalmodel for fitting and predicting relaxation modulus and
simulating viscoelastic responses. Proc R Soc A 2018.
[10]. Zhang Y, Lian Z, Zhou M, Lin T. Viscoelastic behavior of a casing material and its utilization in premium
connections in high-temperature gas wells. Advances in Mechanical Engineering 2018;10(12).
19/05/2021 Tesfaye O. Terefe and Akash Deep 16
 Cont.

The End
Thank you very much

19/05/2021 Tesfaye O. Terefe and Akash Deep 17